Anda di halaman 1dari 12

International Journal of Impact Engineering 22 (1999) 911}922

3D FE-simulation of high-velocity fragment perforation


of reinforced concrete slabs
L. As ga> rdh*, L. Laine
Defence Research Establishment, (FOA), Box 24, S-19 521 MA$ RSTA, Sweden
Received 23 October 1998; received in revised form 17 January 1999

Abstract
In a research project at FOA, various codes and three-dimensional material models for concrete were
used in simulations of some test cases, e.g. impact loads on beams, blast loads from explosives on slabs
and projectile perforation through slabs. Numerical results were compared with observations in tests and
reported. In this paper, simulations using LS-DYNA (LSTC, Livermore, June 1997) for the projectile
perforation of a 60 mm thick "bre-reinforced concrete slab with a velocity of 1500 m/s is presented. The
material model used was Type 78 `Soil/Concretea with erosion. The results were in fairly good agreement
with test results, but more studies are necessary to assess the sensitivity of certain material parameters.
Relevant test procedures and material data for the simulations are discussed in the paper. ( 1999 Elsevier
Science Ltd. All rights reserved.

1. Introduction
Numerical predictions of the structural response to various loads has become an increasingly
helpful tool. Of special interest are predictions of critical conditions like cracking, local and global
failures, perforation by fragments, etc. The choice of software and material models for the used
materials is strongly dependent on the purpose of the computer simulations. For predictions of the
response of reinforced concrete structures to transient dynamic loads from blast and debris,
`Hydrocodesa of type computational #uid dynamics (CFD) and/or "nite element, `explicita
version (FEM) are needed. For the material description, material models suitable for `solid
elementsa, is the most appropriate choice for rapid transient loads, because wave e!ects play an
important role in the stress distribution. Thus small, `cubica elements may be used for modelling
beams and shells.
* Corresponding author.
0734-743X/99/$ - see front matter ( 1999 Elsevier Science Ltd. All rights reserved.
PII: S 0 7 3 4 - 7 4 3 X ( 9 9 ) 0 0 0 0 8 - 1

912

L. As ga> rdh, L. Laine / International Journal of Impact Engineering 22 (1999) 911}922

2. Material models
Concrete models are usually based on a phenomenological behaviour at the macroscopic level.
The hydrostatic and deviatoric stresses are decoupled and represented by tensor invariants. In case
of shock loading the hydrostatic part may be described by the `Hugoniota shock pressure/speci"c
volume relationship (EOS), and the shear strengths are de"ned as yield, failure and residual
strength surfaces in the 3D principal stress space. Normally, the e!ect of internal temperature
changes are neglected in material models for concrete.
The complexity of structural concrete behaviour is basically due to the heterogeneity of the
material. Variations of the properties are due to the production and curing conditions and
interaction with the reinforcement. Thus re"ned concrete models are so far cost e!ective only for
concrete structures produced with extreme accuracy for special purposes, e.g. factory produced
elements, or production under strict surveillance. Thus for numerical simulations it is important to
use a relevant model for every case.
There are several concrete models implemented in LS-DYNA [1], designed for special purposes
as erosion, e!ect of strain rate, cracking, etc. The `Winfrith Concrete Modela, see [3] is a threeinvariant, four-parametric model based on models by Ottosen [4] and Nilsson [5], but with the
strain rate e!ect included as internal enhancements of the uni-axial strengths and the modulus of
elasticity according to recommendations by CEB [6]. The model is advantageous for design
purposes, since the parameters used are based on standardised static tests and since the model is
validated for a number of cases such as blasts and impacts.
New developments and improvements of model Type 16: `Pseudo Tensor Concrete/Geological
Modela are presented by Malvar et al. [7] with the source code. The enhanced model is
implemented in LS-DYNA as Type 72 `Concrete Damagea. A damage model originally developed
at LLNL for the public domain version of DYNA, has been implemented in LS-DYNA as TYPE
96, `Brittle Damage Modela, see [8]. This model also has the advantage of using only a few
material parameters.
In general, the algorithms for the constitutive equations are rather complicated and thus costly
to run. The sensitivity of the material parameters is often unknown to the user, which makes testing
procedures and calculating parameters for the model and the concrete in question an important
issue. Conclusively, successful modelling of concrete structures subjected to fast transient dynamic
loads, requires validated and well-documented material models with parameters that can be
determined from rational test procedures.

3. Material sample test data


Material sample test data is required to calculate the model parameters. Thus material testing
should be performed with great care, following prescribed procedures from taking samples, storing
the samples, performing the tests with calibrated equipment until eventually assessing the resulting
parameters. Unfortunately, this procedure is often abandoned for various reasons. In the design
situation for example, only prescribed quality speci"cations are at hand, not real strength data.
Assessment of an already built structure requires a number of samples from the material `In Situa
as relevant data from the production of the structure is normally lacking.

L. As ga> rdh, L. Laine / International Journal of Impact Engineering 22 (1999) 911}922

913

The standard procedures for strength testing, e.g. cube and cylinder compression tests, manifest
substantial variations besides they do not represent well-de"ned uni-axial compression stress
states. The following procedures for the most important mechanical properties would eventually be
needed for a complete sample testing according to the enhanced models implemented in many
hydrocodes.
f Tri-axial compression tests on 75]150 mm2 cylinders to obtain the equation of state and at
least three points on the compression meridian (Lode angle h"603) and three points on the
tensile meridian (h"03).
f Splitting test on three cylindrical specimens 150]300 mm2 for the estimation of uni-axial
tension strength.
f CEB-method for the estimation of the fracture energy for static loads and possibly also for
transient dynamic loads.
f Split Hopkinson Pressure Bar test to obtain uni-axial stress}strain diagrams for high strain
rates.
f The `Flyer Plate Impact Testa, or similar, to establish Hugoniot shock EOS of importance
for simulating contact detonations and projectile penetration at very high velocities.
See [9].
f Damage parameters, e.g. according to the `Concrete Damagea or `Brittle Damagea models,
Type 72 and 96 or similar for `Soil/Concretea model, Type 78.
Parameters not covered by the procedures given above are the uni-axial compression strength,
the modulus of elasticity and Poisson's number. In case they cannot be estimated based on the
results above with acceptable accuracy, cube or cylinder compression tests have to be performed
according to standard procedures.
In Thigpen and Chin, see [10], a discussion on the determination of "ve phenomenological
coe$cients is presented. A cyclic tri-axial test procedure is needed to obtain the bulk modulus K,
the shear modulus G, the friction coe$cient k, the inelastic volume strain rate/inelastic shear strain
rate b and the hardening coe$cient h. Formulae for the calculations are given.
In general, the required parameters for a speci"ed model without reference to relevant test
procedures or to recommended limiting values, contributes to unnecessary doubts of the achieved
results with that model.

4. Codes and material models used


In this paper results from the perforation simulations with LS-DYNA and the material model
Type 78, `Soil/Concretea are presented. This model was used because of the erosion option that is
useful for this problem.
An equation of state is needed for high pressures. In this case data from tests of concrete
specimens by Gregsson [11], were used, with a uni-axial compressive strength of f "35 MPa and
#
a modulus of elasticity E"34 GPa. Poisson's ratio estimated to the standard value for good
concrete was l"0.2. The solid density for the concrete was reached after a pressure of 12.5 GPa,
giving a remaining plastic volumetric strain e "22%. The concrete used here had f "5 MPa,
74
5

914

L. As ga> rdh, L. Laine / International Journal of Impact Engineering 22 (1999) 911}922

Fig. 1. Shock Hugoniot equation of state, acc. to data from Gregsson [11].

f "80 MPa, E"50 GPa, K"28 GPa, the elastic limit p "26 MPa and e "0.1%. The
#
E
7E
pressure values corresponding to and below e "20% were adjusted according to the data used
7
here, see Fig. 1.
The failure surface, expressed either as the second invariant of the deviator stress tensor J or as
2
the stress di!erence *p"sqrt(3J ), Fig. 2, was computed based on normalised test data published
2
by Balmer, see diagram in e.g. [4]. These test data correspond to the compressive meridian and
reach about 1.08 GPa in the diagram in Fig. 2. The `Three Parametric Polynomiala reaches
a vertex of 1.55 GPa at p"1.8 GPa, after which a horizontal line continues until 10 GPa. As
a comparison a linear approximation of J through (!f , 0) and ( f /3, f 2/3) is included.
2
5
#
#
5. Experiment
In this paper, results from simulations with FE-models are compared with the results from steel
fragment perforation tests reported in [12]. The tests were performed during 1996 with the purpose
of validating numerical models with AUTODYN2D and DYNA3D.
Five steel "bre concrete slabs with the dimensions 400]400]60 mm3, were sawed from square
slabs with originally 1200 mm sides. The slabs were reinforced with a layer of orthogonal bars of
' 8 mm cc 80 mm. Steel quality: Ks400. Cover layer 11 mm. The slab surface closest to the
reinforcement layer is denoted as the `Backa surface. Concrete quality: Bridge deck steel
"bre reinforced concrete, type `Optiroca. Projectiles of steel 2541-3 with '"15 mm, "43 mm
were "red from a powder gas gun with speed around 1500 m/s against the slabs. Velocities were

L. As ga> rdh, L. Laine / International Journal of Impact Engineering 22 (1999) 911}922

915

Fig. 2. Polynomial approximation of Balmers data for the stress di!erence.

Table 1
Results from perforation tests on concrete slabs
Shot slab d

2
3
4
5
6

Velocity in.
(m/s)

1517$5
1524$5
1518$5
1525$5
1526$5

Velocity out
(m/s)

959$30
902$30
935$30
768$30
930$30

Fragment weight (g)


Before

After

60
60
60
60
60

35
45
37
}
}

Notation fragment against

Back surface
-AFront surface
-A-Dev. 503
Front surface

determined before perforation with a laser beam and an electric circuit consisting of two thin
copper sheets on the target and after perforation with two X-ray #ash lights. Test results are shown
in Table 1 and in Figs. 3}8. Shot no 5 deviated ca 503.
In Fig. 3, the fragments are shown after perforation. Fragments from shot nos 5 and 6 were not
found after the test. In Fig. 4, a detail from the backsurface of slab no 3 is shown. Notice that this is
the `penetration surfacea of the target. In Figs. 5 and 6, the back (penetration) and the front (exit)
surfaces of slab no 3 are shown. In Figs. 7 and 8 the front (penetration) and the back (exit) surfaces
of slab no 4 are shown.

916

L. As ga> rdh, L. Laine / International Journal of Impact Engineering 22 (1999) 911}922

Fig. 3. Steel fragments used in the perforation tests. After perforation.

Fig. 4. Detail of the penetration surface of Slab d3 after perforation.

6. Numerical simulations
The FE-model was generated with the pre-processor LS-INGRID ver 34r, see [2]. One quarter
of the slab was modelled, thus two symmetry planes were de"ned. For all the parts included,
eight-node solid elements were used, see Fig. 9. The reinforcement bars were modelled as solid
cylinders with contact surfaces. Static and dynamic friction was de"ned. The elements in the
concrete slab near the perforation area were cubic with 2 mm sides. The total number of elements in
the model was 77 720.
The fragment was a steel cylinder of length 43 mm, diameter 15 mm and weighing 60 g. A quarter
of the cylinder was modelled with 1548 elements. Between the steel fragment and the concrete slab,

L. As ga> rdh, L. Laine / International Journal of Impact Engineering 22 (1999) 911}922

Fig. 5. Penetration surface of Slab d3. Back surface. Crater diam. 210 mm.

Fig. 6. Exit surface of Slab d3. Front surface. Crater diam. 180 mm.

917

918

L. As ga> rdh, L. Laine / International Journal of Impact Engineering 22 (1999) 911}922

Fig. 7. Penetration surface of Slab d4. Front surface. Crater diam. ca 150 mm.

Fig. 8. Exit surface of Slab d4. Back surface. Crater diam. ca 180 mm.

L. As ga> rdh, L. Laine / International Journal of Impact Engineering 22 (1999) 911}922

919

Fig. 9. The FE-model of a steel fragment perforation through the front surface of a steel "bre- and the conventionally
reinforced concrete slab.

Table 2
Failure strain versus pressure. Model d78
Plastic strain

Tryck p"0

p"15 (MPa)

p"100 (MPa)

p"1000 (MPa)

p"10000 (MPa)

e
1
e
2

0
0.01

0
0.01

0.04
0.05

0.49
0.50

4.9
5.0

Fig. 10. Penetration into the target after 15 ls. The initial velocity of 1500 m/s has decreased to 1330 m/s. Shock waves
have reached the rear surface and caused initial spalling.

contact surfaces with erosion were de"ned. The material model, Type 3, `Plastic/Kinematica was
used for the fragment. (Yield strength 400 MPa, Failure strain 30%). The erosion option is based
on the plastic strain. In the concrete model Type 78, erosion occurs when the pressure reaches the
failure pressure limit, see Table 2. The tensile fracture in Type 78 is described with a pressure cut-o!
value, here chosen as !5 MPa.

920

L. As ga> rdh, L. Laine / International Journal of Impact Engineering 22 (1999) 911}922

Fig. 11. The velocity is 1200 m/s after 30 ls. Increased spalling at the rear surface and shear cracks are visible below the
reinforcement bars.

Fig. 12. The velocity is 1050 m/s after 45 ls. Increased erosion in the volume surrounded by reinforcement bars.

7. Results
The simulation of the perforation process used ca 7 h CPU-time on a DEC-Alpha Station 255
4/233. Post processing was performed with LS-Taurus ver. 940. Four phases from the perforation
process are shown in Figs. 10}13 representing time steps impact of 15, 30, 45 and 60 ls after impact.
The striking velocity (at the front surface) was 1500 m/s and the exit velocity 1045 m/s in the
simulations.

L. As ga> rdh, L. Laine / International Journal of Impact Engineering 22 (1999) 911}922

921

Fig. 13. The velocity is 1045 m/s after 60 ls. Most of the volume limited by the reinforcement bars is eroded. Also the
fragment is eroded, compare with Figs. 3 and 8. One can estimate the exit crater to be slightly more than145 mm and the
residual fragment to weigh almost 33 g.

8. Conclusion
Perforation of high velocity objects through structural concrete is one of the most challenging
problems for analysts in forti"cation engineering. The task involves geometrical and material
modelling of a complicated process and, which is a most underestimated task, material sample
testing and determining material parameters.
In the paper, a simulation with the code LS-DYNA and a material model Type 78,
`Soil/Concretea of a perforation test of a 60 mm thick, conventionally and steel "bre reinforced
concrete slab is presented. The fragments were small steel cylinders penetrating with a velocity of ca
1518 m/s and perforating with a rest velocity of ca. 935 m/s. The measured fragment weight was
60 g and the residual weight 37 g. The results from the simulation were 1500 m/s, 1045 m/s and 33 g
respectively. The diameters of the exit craters were measured and found to be around 180 mm in
the tests and 150 mm in the simulation.
The material sample tests performed were: uni-axial compressive strength and splitting tests on
150 mm cubes and the modulus of elasticity on 75]150 mm2 cylinders. Shear strength versus
pressure and pressure versus volumetric strain were derived from sample tests in the literature. The
damage parameters, plastic strain for damage initiation e (p) and completion e (p) versus pressure,
1
2
were estimated after trial and error.
As a result of the simulation, a deeper understanding of the performance of the material model
was obtained. The model as used so far is capable of describing the perforation process qualitatively with an acceptable degree of accuracy, taking into account the absence of complete material
testing procedures, e.g. tri-axial tests for the yield and failure surfaces, `Flyer Plate Impacta-tests
for the shock EOS and the `Split Hopkinson Pressure Bara-tests for the uni-axial dynamic strength
enhancements. Unfortunately, test procedures for the damage development versus (high) pressure
and erosion conditions are not available.

922

L. As ga> rdh, L. Laine / International Journal of Impact Engineering 22 (1999) 911}922

Acknowledgements
The authors wish to express special thanks to Dr. Richard Sturt, Ove Arup and Partners,
London, Prof. Larsgunnar Nilsson, LinkoK ping for guidance and colleagues at FOA who participated in and supported the project.

References
[1] LSTC, LS-DYNA. Keyword User's Manual. Version 940. Non Linear Dynamic Analysis of Structures in Three
Dimensions. Livermore Software Technology Corporation, LSTC, Livermore, June 1997.
[2] LSTC, LS-INGRID. User's Manual. Version 3.3. Livermore Software Technology Corporation, LSTC, Livermore,
June 1997.
[3] Broadhouse BJ, The Winfrith Concrete Model in LS-DYNA3D. Atomic Energy Agency, AEA Technology.
SPD/D(95)363. Dorchester, UK, 1995.
[4] Ottosen NS, Failure and Elasticity of Concrete. Danish Atomic Energy Commission, Research Establishment RisoK ,
RisoK -M-1801, September 1975;67.
[5] Nilsson LG, Impact loading on concrete structures. PhD Dissertation, Chalmers University of Technology, Dept.
of Structural Mechanics, Gothenburgh, 1979.
[6] CEB, Concrete Structures under impact and impulsive loading. Synthesis Report, Bulletin D'Information, No 187,
ComiteH Euro-International du Beton, August 1988.
[7] Malvar LJ, Crawford JE, Weswwich JW, Simons D. A plasticity concrete material model for dyna3D. Int J Impact
Engng 1997; 19(9}10):847}73.
[8] Govindjee S, Gregory JK, Simo JC. Anisotropic modelling and numerical simulation of brittle damage in concrete.
Int. J Numer Analy Methods Geomech 1995;38:3611}33.
[9] Grady D. Shock equation of state properties of concrete. Sandia National Laboratories, Experimental Impact
Physics Department. Sand-95-2215C. Albuquerque, New Mexico, 87185-5800, USA, 1995;10.
[10] Thigpen L, Chin RCY. On constitutive models for compressive loading of geologic materials. UCRL-53073,
Lawrence Livermore National Laboratory, University of California. Livermore Cal. 94550. November 1980;20.
[11] Gregsson VG Jr, A shockwave study of Fondu-Frye WA-1 and a concrete, DNA2797F, February 1992. AFATLTR-78-24-VOL.II, Appendix A, Concrete Constitutive Relations at high Strength Levels. US Army Armament
Research and Development Command, Aberdeen, March, USA, 1978.
[12] Bryntse A, Penetration av sta> lsplitter vid 1500 m/s mot armerad betong. Defence Research Establishment,
FOA-D-97-00304-311}SE. Stockholm, January 1997.

Anda mungkin juga menyukai